I think I spoke too soon! The debugging is very difficult because I have no symbol tables nor source code.
Stephen Brooks sent me a debug version to try. It generates a GP fault with a floating point stack fault. Not exactly what I'm getting with the current version. It's in the same routine, though. I think that it's trying to create a temporary file and the rand() routine fails somehow.
I'll be working on it all weekend and rabrittain will be helping me, since I don't have XP x86 to run a parallel on.
If the x86 version works on his machine, I'll have Rex apply his FX-60 to DPAD, which will help a little.